A Distributed Architecture for Simulation Environments Based on Game Engine Systems - Visual Computing: Scientific Visualization and Imaging Systems - page 57

Each computer have an Automatic Update Task that receives task from the
Distribution Task and executes the task, when it finishes the task execution it sends
the results back to the server. If a computer loses the link with the server, the
server will automatic distribute its task to another client.
2.3.4 The Configuration Script
The configuration script is used in order to configure how the automatic distri-
bution task will execute the heuristic. This script can be used to define:
• the maximum numbers of computers that will be used in the simulation;
• the computers IP addresses;
• the network mode (if it will run the tasks locally or in the network distribution
mode);
• the tasks that can use the network mode;
• the minimum network connection speed between the server and the client in
order to use that client machine.
2.3.5 A Summary of the Architecture Execution
First, the server computer (which displays visualization and gathers the player
input) queries the network for available computers and their capacity. After that, it
divides the amount of work between the computers, considering the processing
power of each one. While running the game loop, the server verifies if any
computer is down or if there are lost connections, in this case, it tries to redistribute
the task to the other remaining computers. It also exchange messages with the
computers, process its own tasks that it might have been assigned to, and then
presents the results to the end user.
2.4 Conclusion
With the evolution of computer networks, distributing computation will become
more in evidence, allowing more complex real-time simulations.
This work presents and summarizes the concept of the main loops present in
simulations, a subject that is not very much discussed in the literature. We discuss
an architecture for loops that are able to distribute tasks between computers in a
network, and inside computers through CPUs, CPU cores, and GPUs. With this
approach a simulator is able to use more resources available to it (local and
remote), reducing its system requirements.